I don't think this blogger actually read the post by Google. The neosmart article emphasizes This is a departure from the process outlined in OpenID 1.0. This is because Google is following the OpenID 2.0 protocol which does contain the discovery request and the XRDS document process.
According to The Next Web (see this post: http://cli.gs/MM0aab ), "Dick Hardt, an important OpenID supporter, also confirms that Google is NOT compatible with OpenID". They link to his comment on TC.
http://openid.net/specs/openid-authentication-2_0.html#disco...
This is not a fork from OpenID. They are simply using the newer protocol.